Twitxy Saved

Michael Choi

@michaelchchoi

Assistant Professor @NUSingapore. Applied probabilist. Probability, MCMC, statistical physics, optimization, information theory, TCS. Opinions my own.
Singapore mchchoi.github.io Joined October 2015
3.1K Following 3.4K Followers 2.1K Posts

Latest posts 5

Pinned post

Michael Choi @michaelchchoi

Replying to @michaelchchoi

[2/2]
2,3. Group-averaged Markov chains I & II (with Youjia Wang, Ryan Lim)
I: arxiv.org/abs/2509.02996
II: arxiv.org/abs/2512.13067
Comments: We generalize equi-probability permutations to group actions—beautiful blend of Markov chain theory, info theory, and group theory.
arxiv.org Group-averaged Markov chains: mixing improvement For Markov kernels $P$ on a general state space $\mathcal{X}$, we introduce a new class of averaged Markov kernels $P_{da}(G,ν)$ of $P$ induced by a group $G$ that acts on $\mathcal{X}$ and a...
1 replies 4 reposts 37 likes 14K views View on X
Michael Choi @michaelchchoi
[1/3] Consider a Markowitz mean-variance problem but for MCMC. Suppose we have a baseline sampler ("assets") G_0 = P along with some other (Gibbs) samplers G_1, G_2. How to allocate the weights alpha_i for i = 0,1,2 such that 1. the expected diversity ("return") is maximized,
Michael Choi @michaelchchoi
In December 2022, I visited ISM and RIKEN AIP. I met Geoff Wolfer where I learnt the basics of Information Geometry (IG) @SN_INGE. This eventually kickstarted my current research program on using IG (particularly information projections) to accelerate MCMC samplers.
0 replies 0 reposts 10 likes 658 views View on X